Linux

In the Linux kernel, the following vulnerability has been resolved: netfilter: ipset: Fix race between namespace cleanup and gc in the list:set type Lion Ackermann reported that there is a race condition between namespace cleanup in ipset and the garbage collection of the list:set type. The namespace cleanup can destroy the list:set type of sets while the gc of the set type is waiting to run in rcu cleanup. The latter uses data from the destroyed set which thus leads use after free. The patch contains the following parts: - When destroying all sets, first remove the garbage collectors, then wait if needed and then destroy the sets. - Fix the badly ordered "wait then remove gc" for the destroy a single set case. - Fix the missing rcu locking in the list:set type in the userspace test case. - Use proper RCU list handlings in the list:set type. The patch depends on c1193d9bbbd3 (netfilter: ipset: Add list flush to cancel_gc).Show less

In the Linux kernel, the following vulnerability has been resolved: ionic: fix use after netif_napi_del() When queues are started, netif_napi_add() and napi_enable() are called. If there are 4 queues and only 3 queues are used for the current configuration, only 3 queues' napi should be registered and enabled. The ionic_qcq_enable() checks whether the .poll pointer is not NULL for enabling only the using queue' napi. Unused queues' napi will not be registered by netif_napi_add(), so the .poll pointer indicates NULL. But it couldn't distinguish whether the napi was unregistered or not because netif_napi_del() doesn't reset the .poll pointer to NULL. So, ionic_qcq_enable() calls napi_enable() for the queue, which was unregistered by netif_napi_del(). Reproducer: ethtool -L <interface name> rx 1 tx 1 combined 0 ethtool -L <interface name> rx 0 tx 0 combined 1 ethtool -L <interface name> rx 0 tx 0 combined 4 Splat looks like: kernel BUG at net/core/dev.c:6666! Oops: invalid opcode: 0000 [#1] PREEMPT SMP NOPTI CPU: 3 PID: 1057 Comm: kworker/3:3 Not tainted 6.10.0-rc2+ #16 Workqueue: events ionic_lif_deferred_work [ionic] RIP: 0010:napi_enable+0x3b/0x40 Code: 48 89 c2 48 83 e2 f6 80 b9 61 09 00 00 00 74 0d 48 83 bf 60 01 00 00 00 74 03 80 ce 01 f0 4f RSP: 0018:ffffb6ed83227d48 EFLAGS: 00010246 RAX: 0000000000000000 RBX: ffff97560cda0828 RCX: 0000000000000029 RDX: 0000000000000001 RSI: 0000000000000000 RDI: ffff97560cda0a28 RBP: ffffb6ed83227d50 R08: 0000000000000400 R09: 0000000000000001 R10: 0000000000000001 R11: 0000000000000001 R12: 0000000000000000 R13: ffff97560ce3c1a0 R14: 0000000000000000 R15: ffff975613ba0a20 FS: 0000000000000000(0000) GS:ffff975d5f780000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00007f8f734ee200 CR3: 0000000103e50000 CR4: 00000000007506f0 PKRU: 55555554 Call Trace: <TASK> ? die+0x33/0x90 ? do_trap+0xd9/0x100 ? napi_enable+0x3b/0x40 ? do_error_trap+0x83/0xb0 ? napi_enable+0x3b/0x40 ? napi_enable+0x3b/0x40 ? exc_invalid_op+0x4e/0x70 ? napi_enable+0x3b/0x40 ? asm_exc_invalid_op+0x16/0x20 ? napi_enable+0x3b/0x40 ionic_qcq_enable+0xb7/0x180 [ionic 59bdfc8a035436e1c4224ff7d10789e3f14643f8] ionic_start_queues+0xc4/0x290 [ionic 59bdfc8a035436e1c4224ff7d10789e3f14643f8] ionic_link_status_check+0x11c/0x170 [ionic 59bdfc8a035436e1c4224ff7d10789e3f14643f8] ionic_lif_deferred_work+0x129/0x280 [ionic 59bdfc8a035436e1c4224ff7d10789e3f14643f8] process_one_work+0x145/0x360 worker_thread+0x2bb/0x3d0 ? __pfx_worker_thread+0x10/0x10 kthread+0xcc/0x100 ? __pfx_kthread+0x10/0x10 ret_from_fork+0x2d/0x50 ? __pfx_kthread+0x10/0x10 ret_from_fork_asm+0x1a/0x30Show less

In the Linux kernel, the following vulnerability has been resolved: sock_map: avoid race between sock_map_close and sk_psock_put sk_psock_get will return NULL if the refcount of psock has gone to 0, which will happen when the last call of sk_psock_put is done. However, sk_psock_drop may not have finished yet, so the close callback will still point to sock_map_close despite psock being NULL. This can be reproduced with a thread deleting an element from the sock map, while the second one creates a socket, adds it to the map and closes it. That will trigger the WARN_ON_ONCE: ------------[ cut here ]------------ WARNING: CPU: 1 PID: 7220 at net/core/sock_map.c:1701 sock_map_close+0x2a2/0x2d0 net/core/sock_map.c:1701 Modules linked in: CPU: 1 PID: 7220 Comm: syz-executor380 Not tainted 6.9.0-syzkaller-07726-g3c999d1ae3c7 #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 04/02/2024 RIP: 0010:sock_map_close+0x2a2/0x2d0 net/core/sock_map.c:1701 Code: df e8 92 29 88 f8 48 8b 1b 48 89 d8 48 c1 e8 03 42 80 3c 20 00 74 08 48 89 df e8 79 29 88 f8 4c 8b 23 eb 89 e8 4f 15 23 f8 90 <0f> 0b 90 48 83 c4 08 5b 41 5c 41 5d 41 5e 41 5f 5d e9 13 26 3d 02 RSP: 0018:ffffc9000441fda8 EFLAGS: 00010293 RAX: ffffffff89731ae1 RBX: ffffffff94b87540 RCX: ffff888029470000 RDX: 0000000000000000 RSI: ffffffff8bcab5c0 RDI: ffffffff8c1faba0 RBP: 0000000000000000 R08: ffffffff92f9b61f R09: 1ffffffff25f36c3 R10: dffffc0000000000 R11: fffffbfff25f36c4 R12: ffffffff89731840 R13: ffff88804b587000 R14: ffff88804b587000 R15: ffffffff89731870 FS: 000055555e080380(0000) GS:ffff8880b9500000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000000 CR3: 00000000207d4000 CR4: 0000000000350ef0 Call Trace: <TASK> unix_release+0x87/0xc0 net/unix/af_unix.c:1048 __sock_release net/socket.c:659 [inline] sock_close+0xbe/0x240 net/socket.c:1421 __fput+0x42b/0x8a0 fs/file_table.c:422 __do_sys_close fs/open.c:1556 [inline] __se_sys_close fs/open.c:1541 [inline] __x64_sys_close+0x7f/0x110 fs/open.c:1541 do_syscall_x64 arch/x86/entry/common.c:52 [inline] do_syscall_64+0xf5/0x240 arch/x86/entry/common.c:83 entry_SYSCALL_64_after_hwframe+0x77/0x7f RIP: 0033:0x7fb37d618070 Code: 00 00 48 c7 c2 b8 ff ff ff f7 d8 64 89 02 b8 ff ff ff ff eb d4 e8 10 2c 00 00 80 3d 31 f0 07 00 00 74 17 b8 03 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 48 c3 0f 1f 80 00 00 00 00 48 83 ec 18 89 7c RSP: 002b:00007ffcd4a525d8 EFLAGS: 00000202 ORIG_RAX: 0000000000000003 RAX: ffffffffffffffda RBX: 0000000000000005 RCX: 00007fb37d618070 RDX: 0000000000000010 RSI: 00000000200001c0 RDI: 0000000000000004 RBP: 0000000000000000 R08: 0000000100000000 R09: 0000000100000000 R10: 0000000000000000 R11: 0000000000000202 R12: 0000000000000000 R13: 0000000000000000 R14: 0000000000000000 R15: 0000000000000000 </TASK> Use sk_psock, which will only check that the pointer is not been set to NULL yet, which should only happen after the callbacks are restored. If, then, a reference can still be gotten, we may call sk_psock_stop and cancel psock->work. As suggested by Paolo Abeni, reorder the condition so the control flow is less convoluted. After that change, the reproducer does not trigger the WARN_ON_ONCE anymore.Show less

In the Linux kernel, the following vulnerability has been resolved: vmci: prevent speculation leaks by sanitizing event in event_deliver() Coverity spotted that event_msg is controlled by user-space, event_msg->event_data.event is passed to event_deliver() and used as an index without sanitization. This change ensures that the event index is sanitized to mitigate any possibility of speculative information leaks. This bug was discovered and resolved using Coverity Static Analysis Security Testing (SAST) by Synopsys, Inc. Only compile tested, no access to HW.Show less

In the Linux kernel, the following vulnerability has been resolved: drm/mst: Fix NULL pointer dereference at drm_dp_add_payload_part2 [Why] Commit: - commit 5aa1dfcdf0a4 ("drm/mst: Refactor the flow for payload allocation/removement") accidently overwrite the commit - commit 54d217406afe ("drm: use mgr->dev in drm_dbg_kms in drm_dp_add_payload_part2") which cause regression. [How] Recover the original NULL fix and remove the unnecessary input parameter 'state' for drm_dp_add_payload_part2(). (cherry picked from commit 4545614c1d8da603e57b60dd66224d81b6ffc305)Show less

In the Linux kernel, the following vulnerability has been resolved: drm/shmem-helper: Fix BUG_ON() on mmap(PROT_WRITE, MAP_PRIVATE) Lack of check for copy-on-write (COW) mapping in drm_gem_shmem_mmap allows users to call mmap with PROT_WRITE and MAP_PRIVATE flag causing a kernel panic due to BUG_ON in vmf_insert_pfn_prot: BUG_ON((vma->vm_flags & VM_PFNMAP) && is_cow_mapping(vma->vm_flags)); Return -EINVAL early if COW mapping is detected. This bug affects all drm drivers using default shmem helpers. It can be reproduced by this simple example: void *ptr = mmap(0, size, PROT_WRITE, MAP_PRIVATE, fd, mmap_offset); ptr[0] = 0;Show less

In the Linux kernel, the following vulnerability has been resolved: btrfs: zoned: fix use-after-free due to race with dev replace While loading a zone's info during creation of a block group, we can race with a device replace operation and then trigger a use-after-free on the device that was just replaced (source device of the replace operation). This happens because at btrfs_load_zone_info() we extract a device from the chunk map into a local variable and then use the device while not under the protection of the device replace rwsem. So if there's a device replace operation happening when we extract the device and that device is the source of the replace operation, we will trigger a use-after-free if before we finish using the device the replace operation finishes and frees the device. Fix this by enlarging the critical section under the protection of the device replace rwsem so that all uses of the device are done inside the critical section.Show less

In the Linux kernel, the following vulnerability has been resolved: greybus: Fix use-after-free bug in gb_interface_release due to race condition. In gb_interface_create, &intf->mode_switch_completion is bound with gb_interface_mode_switch_work. Then it will be started by gb_interface_request_mode_switch. Here is the relevant code. if (!queue_work(system_long_wq, &intf->mode_switch_work)) { ... } If we call gb_interface_release to make cleanup, there may be an unfinished work. This function will call kfree to free the object "intf". However, if gb_interface_mode_switch_work is scheduled to run after kfree, it may cause use-after-free error as gb_interface_mode_switch_work will use the object "intf". The possible execution flow that may lead to the issue is as follows: CPU0 CPU1 | gb_interface_create | gb_interface_request_mode_switch gb_interface_release | kfree(intf) (free) | | gb_interface_mode_switch_work | mutex_lock(&intf->mutex) (use) Fix it by canceling the work before kfree.Show less

In the Linux kernel, the following vulnerability has been resolved: ima: Fix use-after-free on a dentry's dname.name ->d_name.name can change on rename and the earlier value can be freed; there are conditions sufficient to stabilize it (->d_lock on dentry, ->d_lock on its parent, ->i_rwsem exclusive on the parent's inode, rename_lock), but none of those are met at any of the sites. Take a stable snapshot of the name instead.Show less

In the Linux kernel, the following vulnerability has been resolved: crypto: qat - Fix ADF_DEV_RESET_SYNC memory leak Using completion_done to determine whether the caller has gone away only works after a complete call. Furthermore it's still possible that the caller has not yet called wait_for_completion, resulting in another potential UAF. Fix this by making the caller use cancel_work_sync and then freeing the memory safely.Show less

In the Linux kernel, the following vulnerability has been resolved: mailbox: mtk-cmdq: Fix pm_runtime_get_sync() warning in mbox shutdown The return value of pm_runtime_get_sync() in cmdq_mbox_shutdown() will return 1 when pm runtime state is active, and we don't want to get the warning message in this case. So we change the return value < 0 for WARN_ON().Show less

In the Linux kernel, the following vulnerability has been resolved: ALSA: hda: cs35l56: Fix lifetime of cs_dsp instance The cs_dsp instance is initialized in the driver probe() so it should be freed in the driver remove(). Also fix a missing call to cs_dsp_remove() in the error path of cs35l56_hda_common_probe(). The call to cs_dsp_remove() was being done in the component unbind callback cs35l56_hda_unbind(). This meant that if the driver was unbound and then re-bound it would be using an uninitialized cs_dsp instance. It is best to initialize the cs_dsp instance in probe() so that it can return an error if it fails. The component binding API doesn't have any error handling so there's no way to handle a failure if cs_dsp was initialized in the bind.Show less

In the Linux kernel, the following vulnerability has been resolved: ipv6: sr: fix missing sk_buff release in seg6_input_core The seg6_input() function is responsible for adding the SRH into a packet, delegating the operation to the seg6_input_core(). This function uses the skb_cow_head() to ensure that there is sufficient headroom in the sk_buff for accommodating the link-layer header. In the event that the skb_cow_header() function fails, the seg6_input_core() catches the error but it does not release the sk_buff, which will result in a memory leak. This issue was introduced in commit af3b5158b89d ("ipv6: sr: fix BUG due to headroom too small after SRH push") and persists even after commit 7a3f5b0de364 ("netfilter: add netfilter hooks to SRv6 data plane"), where the entire seg6_input() code was refactored to deal with netfilter hooks. The proposed patch addresses the identified memory leak by requiring the seg6_input_core() function to release the sk_buff in the event that skb_cow_head() fails.Show less

In the Linux kernel, the following vulnerability has been resolved: ipv6: sr: fix memleak in seg6_hmac_init_algo seg6_hmac_init_algo returns without cleaning up the previous allocations if one fails, so it's going to leak all that memory and the crypto tfms. Update seg6_hmac_exit to only free the memory when allocated, so we can reuse the code directly.Show less

In the Linux kernel, the following vulnerability has been resolved: arm64: asm-bug: Add .align 2 to the end of __BUG_ENTRY When CONFIG_DEBUG_BUGVERBOSE=n, we fail to add necessary padding bytes to bug_table entries, and as a result the last entry in a bug table will be ignored, potentially leading to an unexpected panic(). All prior entries in the table will be handled correctly. The arm64 ABI requires that struct fields of up to 8 bytes are naturally-aligned, with padding added within a struct such that struct are suitably aligned within arrays. When CONFIG_DEBUG_BUGVERPOSE=y, the layout of a bug_entry is: struct bug_entry { signed int bug_addr_disp; // 4 bytes signed int file_disp; // 4 bytes unsigned short line; // 2 bytes unsigned short flags; // 2 bytes } ... with 12 bytes total, requiring 4-byte alignment. When CONFIG_DEBUG_BUGVERBOSE=n, the layout of a bug_entry is: struct bug_entry { signed int bug_addr_disp; // 4 bytes unsigned short flags; // 2 bytes < implicit padding > // 2 bytes } ... with 8 bytes total, with 6 bytes of data and 2 bytes of trailing padding, requiring 4-byte alginment. When we create a bug_entry in assembly, we align the start of the entry to 4 bytes, which implicitly handles padding for any prior entries. However, we do not align the end of the entry, and so when CONFIG_DEBUG_BUGVERBOSE=n, the final entry lacks the trailing padding bytes. For the main kernel image this is not a problem as find_bug() doesn't depend on the trailing padding bytes when searching for entries: for (bug = __start___bug_table; bug < __stop___bug_table; ++bug) if (bugaddr == bug_addr(bug)) return bug; However for modules, module_bug_finalize() depends on the trailing bytes when calculating the number of entries: mod->num_bugs = sechdrs[i].sh_size / sizeof(struct bug_entry); ... and as the last bug_entry lacks the necessary padding bytes, this entry will not be counted, e.g. in the case of a single entry: sechdrs[i].sh_size == 6 sizeof(struct bug_entry) == 8; sechdrs[i].sh_size / sizeof(struct bug_entry) == 0; Consequently module_find_bug() will miss the last bug_entry when it does: for (i = 0; i < mod->num_bugs; ++i, ++bug) if (bugaddr == bug_addr(bug)) goto out; ... which can lead to a kenrel panic due to an unhandled bug. This can be demonstrated with the following module: static int __init buginit(void) { WARN(1, "hello\n"); return 0; } static void __exit bugexit(void) { } module_init(buginit); module_exit(bugexit); MODULE_LICENSE("GPL"); ... which will trigger a kernel panic when loaded: ------------[ cut here ]------------ hello Unexpected kernel BRK exception at EL1 Internal error: BRK handler: 00000000f2000800 [#1] PREEMPT SMP Modules linked in: hello(O+) CPU: 0 PID: 50 Comm: insmod Tainted: G O 6.9.1 #8 Hardware name: linux,dummy-virt (DT) pstate: 60400005 (nZCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--) pc : buginit+0x18/0x1000 [hello] lr : buginit+0x18/0x1000 [hello] sp : ffff800080533ae0 x29: ffff800080533ae0 x28: 0000000000000000 x27: 0000000000000000 x26: ffffaba8c4e70510 x25: ffff800080533c30 x24: ffffaba8c4a28a58 x23: 0000000000000000 x22: 0000000000000000 x21: ffff3947c0eab3c0 x20: ffffaba8c4e3f000 x19: ffffaba846464000 x18: 0000000000000006 x17: 0000000000000000 x16: ffffaba8c2492834 x15: 0720072007200720 x14: 0720072007200720 x13: ffffaba8c49b27c8 x12: 0000000000000312 x11: 0000000000000106 x10: ffffaba8c4a0a7c8 x9 : ffffaba8c49b27c8 x8 : 00000000ffffefff x7 : ffffaba8c4a0a7c8 x6 : 80000000fffff000 x5 : 0000000000000107 x4 : 0000000000000000 x3 : 0000000000000000 x2 : 0000000000000000 x1 : 0000000000000000 x0 : ffff3947c0eab3c0 Call trace: buginit+0x18/0x1000 [hello] do_one_initcall+0x80/0x1c8 do_init_module+0x60/0x218 load_module+0x1ba4/0x1d70 __do_sys_init_module+0x198/0x1d0 __arm64_sys_init_module+0x1c/0x28 invoke_syscall+0x48/0x114 el0_svc ---truncated---Show less

In the Linux kernel, the following vulnerability has been resolved: bonding: Fix out-of-bounds read in bond_option_arp_ip_targets_set() In function bond_option_arp_ip_targets_set(), if newval->string is an empty string, newval->string+1 will point to the byte after the string, causing an out-of-bound read. BUG: KASAN: slab-out-of-bounds in strlen+0x7d/0xa0 lib/string.c:418 Read of size 1 at addr ffff8881119c4781 by task syz-executor665/8107 CPU: 1 PID: 8107 Comm: syz-executor665 Not tainted 6.7.0-rc7 #1 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.15.0-1 04/01/2014 Call Trace: <TASK> __dump_stack lib/dump_stack.c:88 [inline] dump_stack_lvl+0xd9/0x150 lib/dump_stack.c:106 print_address_description mm/kasan/report.c:364 [inline] print_report+0xc1/0x5e0 mm/kasan/report.c:475 kasan_report+0xbe/0xf0 mm/kasan/report.c:588 strlen+0x7d/0xa0 lib/string.c:418 __fortify_strlen include/linux/fortify-string.h:210 [inline] in4_pton+0xa3/0x3f0 net/core/utils.c:130 bond_option_arp_ip_targets_set+0xc2/0x910 drivers/net/bonding/bond_options.c:1201 __bond_opt_set+0x2a4/0x1030 drivers/net/bonding/bond_options.c:767 __bond_opt_set_notify+0x48/0x150 drivers/net/bonding/bond_options.c:792 bond_opt_tryset_rtnl+0xda/0x160 drivers/net/bonding/bond_options.c:817 bonding_sysfs_store_option+0xa1/0x120 drivers/net/bonding/bond_sysfs.c:156 dev_attr_store+0x54/0x80 drivers/base/core.c:2366 sysfs_kf_write+0x114/0x170 fs/sysfs/file.c:136 kernfs_fop_write_iter+0x337/0x500 fs/kernfs/file.c:334 call_write_iter include/linux/fs.h:2020 [inline] new_sync_write fs/read_write.c:491 [inline] vfs_write+0x96a/0xd80 fs/read_write.c:584 ksys_write+0x122/0x250 fs/read_write.c:637 do_syscall_x64 arch/x86/entry/common.c:52 [inline] do_syscall_64+0x40/0x110 arch/x86/entry/common.c:83 entry_SYSCALL_64_after_hwframe+0x63/0x6b ---[ end trace ]--- Fix it by adding a check of string length before using it.Show less

In the Linux kernel, the following vulnerability has been resolved: drm/drm_file: Fix pid refcounting race <maarten.lankhorst@linux.intel.com>, Maxime Ripard <mripard@kernel.org>, Thomas Zimmermann <tzimmermann@suse.de> filp->pid is supposed to be a refcounted pointer; however, before this patch, drm_file_update_pid() only increments the refcount of a struct pid after storing a pointer to it in filp->pid and dropping the dev->filelist_mutex, making the following race possible: process A process B ========= ========= begin drm_file_update_pid mutex_lock(&dev->filelist_mutex) rcu_replace_pointer(filp->pid, <pid B>, 1) mutex_unlock(&dev->filelist_mutex) begin drm_file_update_pid mutex_lock(&dev->filelist_mutex) rcu_replace_pointer(filp->pid, <pid A>, 1) mutex_unlock(&dev->filelist_mutex) get_pid(<pid A>) synchronize_rcu() put_pid(<pid B>) *** pid B reaches refcount 0 and is freed here *** get_pid(<pid B>) *** UAF *** synchronize_rcu() put_pid(<pid A>) As far as I know, this race can only occur with CONFIG_PREEMPT_RCU=y because it requires RCU to detect a quiescent state in code that is not explicitly calling into the scheduler. This race leads to use-after-free of a "struct pid". It is probably somewhat hard to hit because process A has to pass through a synchronize_rcu() operation while process B is between mutex_unlock() and get_pid(). Fix it by ensuring that by the time a pointer to the current task's pid is stored in the file, an extra reference to the pid has been taken. This fix also removes the condition for synchronize_rcu(); I think that optimization is unnecessary complexity, since in that case we would usually have bailed out on the lockless check above.Show less

In the Linux kernel, the following vulnerability has been resolved: media: v4l: async: Properly re-initialise notifier entry in unregister The notifier_entry of a notifier is not re-initialised after unregistering the notifier. This leads to dangling pointers being left there so use list_del_init() to return the notifier_entry an empty list.Show less

In the Linux kernel, the following vulnerability has been resolved: mmc: davinci: Don't strip remove function when driver is builtin Using __exit for the remove function results in the remove callback being discarded with CONFIG_MMC_DAVINCI=y. When such a device gets unbound (e.g. using sysfs or hotplug), the driver is just removed without the cleanup being performed. This results in resource leaks. Fix it by compiling in the remove callback unconditionally. This also fixes a W=1 modpost warning: WARNING: modpost: drivers/mmc/host/davinci_mmc: section mismatch in reference: davinci_mmcsd_driver+0x10 (section: .data) -> davinci_mmcsd_remove (section: .exit.text)Show less

In the Linux kernel, the following vulnerability has been resolved: KVM: SVM: WARN on vNMI + NMI window iff NMIs are outright masked When requesting an NMI window, WARN on vNMI support being enabled if and only if NMIs are actually masked, i.e. if the vCPU is already handling an NMI. KVM's ABI for NMIs that arrive simultanesouly (from KVM's point of view) is to inject one NMI and pend the other. When using vNMI, KVM pends the second NMI simply by setting V_NMI_PENDING, and lets the CPU do the rest (hardware automatically sets V_NMI_BLOCKING when an NMI is injected). However, if KVM can't immediately inject an NMI, e.g. because the vCPU is in an STI shadow or is running with GIF=0, then KVM will request an NMI window and trigger the WARN (but still function correctly). Whether or not the GIF=0 case makes sense is debatable, as the intent of KVM's behavior is to provide functionality that is as close to real hardware as possible. E.g. if two NMIs are sent in quick succession, the probability of both NMIs arriving in an STI shadow is infinitesimally low on real hardware, but significantly larger in a virtual environment, e.g. if the vCPU is preempted in the STI shadow. For GIF=0, the argument isn't as clear cut, because the window where two NMIs can collide is much larger in bare metal (though still small). That said, KVM should not have divergent behavior for the GIF=0 case based on whether or not vNMI support is enabled. And KVM has allowed simultaneous NMIs with GIF=0 for over a decade, since commit 7460fb4a3400 ("KVM: Fix simultaneous NMIs"). I.e. KVM's GIF=0 handling shouldn't be modified without a *really* good reason to do so, and if KVM's behavior were to be modified, it should be done irrespective of vNMI support.Show less